Mavju AI
TEAM
THE PROJECT GOAL
Sphere: Banking
The problem
You are a farmer or a small business owner
- You work informally
- No accounting
- Income is unstable
- Everything runs through cash or USSD
You need money
to grow
- Buy seeds or fertilizer
- Purchase goods
- Repair equipment
- Expand operations
You try to get a loan from the bank
- Fill out forms
- Provide documents you
often don’t have
- Wait for manual review
The bank demands collateral or credit history
- Land title
- Property
- Deposit
- Official income records
Your application gets rejected
- 70% of SME borrowers are declined
- Not because they can’t repay
- But because the system can’t evaluate them
You turn to informal lenders
- 20–30% interest per month
- Debt trap
- You lose most of your income
The solution
You enter your basic business information
- type of business
- monthly income
- suppliers and buyers
- payment or USSD history
AI analyzes your real economic behavior
- consistency of income
- stability of supplier/buyer relationships
- cash-flow patterns
- risk indicators
Instant credit decision
- Approved / Declined / Conditional
- Recommended credit limit
- Fair interest rate
- Explanation of the decision
You receive your credit and grow your business
- You get funding with fair interest
- Invest into goods, equipment, or expansion.
- Your business grows sustainably instead of falling into debt.
WHY OUR TEAM CAN SOLVE THIS
We are a focused two-person team combining full-stack engineering, AI/ML and research for financial inclusion in Uzbekistan.
Muhammad builds reliable, production-ready systems end-to-end. Ruslan designs and validates AI models and brings deep context on SME and informal lending in Central Asia.
We move fast, keep the scope realistic for a hackathon and are personally motivated to reduce 20–30% monthly informal interest rates that destroy small businesses.
✔ Full-stack + AI/ML skillset
✔ Understanding of local SME / farmer reality
✔ Experience with data, APIs and UX
✔ Strong motivation to build a real product, not a demo
ROADMAP & CURRENT STAGE
Current stage: Idea
Roadmap
- Week 1–2: Synthetic dataset & baseline ML model
- Week 3: Scoring API (FastAPI backend)
- Week 4: Simple web UI + API integration
- Week 5: Fairness checks & explainability (SHAP-based)
- Week 6: Model improvement, stress-testing and bank-ready version
Next steps after hackathon
- Connect real alternative data sources (USSD, suppliers, payments)
- Validate scoring quality together with a banking partner (e.g. Agrobank)
- Deploy pilot for a limited group of SME borrowers and farmers
- Iterate on accuracy, fairness and UX based on real-world feedback
HOW WE PLAN TO IMPLEMENT THE SOLUTION
Approach
- Collect and generate synthetic data that imitates SME / farmer behavior
- Engineer behavioral features: income stability, supplier/buyer network, payment discipline, risk patterns
- Train a machine learning model (XGBoost) to estimate probability of default
- Wrap the model in a credit-scoring API and connect a minimal web interface
- Add explainability (SHAP) and fairness checks for gender / region
Tech stack & AI
- Frontend: HTML, CSS, JavaScript, Vue.js
- Backend: Python, FastAPI
- ML: XGBoost, Scikit-Learn, Pandas, synthetic data generation
- AI usage: behavioral credit scoring, risk prediction, fairness constraints
- Explainable AI: SHAP values to show why a borrower is approved or rejected